Big Data Analytics MCQs

Futures allow nonblocking composition of asynchronous computations.
An execution context schedules the underlying work.

Laziness allows Spark to optimize and pipeline operations.
Unused transformation chains do not perform work.

collect is safe only when the complete result fits in driver memory.
Large results can crash or overload the driver.

Persistence avoids recomputing reused RDDs across actions.
Storage levels can use memory, disk, or serialized forms.

Projection chooses and computes output columns.
It is analogous to the SELECT list in SQL.

Many machine-learning algorithms require numeric representations.
The fitted indexer preserves a consistent category mapping.

An ExecutionContext provides threads or an execution service.
Poorly chosen contexts can cause starvation or excessive concurrency.

Lineage is a logical dependency graph.
Spark uses it for planning and fault recovery.

take limits driver-side data volume.
It is useful for checking records without retrieving the whole RDD.

cache is shorthand for a default persist choice.
Actual materialization occurs when an action evaluates the RDD.

Filtering reduces the row set.
Spark may push supported predicates down to the data source.

One-hot encoding avoids imposing arbitrary numeric order on categories.
Sparse storage is efficient when there are many levels.